On Sun, 20 Jan 2008 06:40:03 -0800, David wrote:

Did you visit the ESE web site? They make the master clocks used in
broadcasting (and a lot of public service and labs) and they have very
well thought-out innards.

Yes, and yes. I never meant to insinuate that it was impossible to
obtain an accurate time standard using GPS. I did mean to point out
that to do so requires a properly designed receiver, which goes way
beyond the type of consumer-grade device the original poster had in
mind when telling us about NIST's plans to possibly add an east coast
equivalent to WWVB.

Is GPS like NTP in as they just give a raw number of seconds since a
certain date and that the receiver computes the real time from that,
based on its firmware?

Well...I'd have to say yes, since they are both based on the same
standard, which is TAI.

As I understand things, NTP actually references the UTC time scale
with a refined version of Marzullo's Algorithm called the intersection
algorithm. Like the original Marzullo's algorithm, the intersection
algorithm is used to select from among several sources for estimating
accurate time. It supports leap seconds, and the sources it uses can
include atomic clocks, GPS clocks or other radio clocks, computers
that are directly connected to such devices, and/or computers that are
able to use the Internet to poll computers that are directly connected
to such devices.

UTC (again as I understand things) is defined by a large collection
(about 300) of atomic clocks in over 50 national laboratories
worldwide, using the atomic time scale TAI (Temps Atomique
International, or International Atomic Time). TAI is sort of based on
counting seconds, in that it counts something called SI seconds, which
are defined in relationship to the radiation state of the Cesium-133
atom by international agreement. UTC is then derived from TAI, and is
periodically updated with leap seconds in order to more closely track
time standards based on the Earth's rotation.

I think I pretty much explained how GPS receivers estimate the time in
my previous reply upthread to a post by Geoffrey, so I won't repeat
all that again here, but when all is said and done, a fixed 19 second
offset from TAI gives GPS time. UTC and TAI were synchronized to an
exact fixed offset of 33 seconds at the end of 1971.

So the answer to your question. I would conclude that GPS and NTP are
alike in that both are synchronized to the same international
standard, based on cesium atomic clocks.

JK